Fragmented workflows
Requirements, simulation scripts, artifacts, and engineering discussions can become difficult to follow across tools.
Emerging research-to-product initiative / Analog EDA + AI
Analog design.
Connected thinking.
Meet ANVIL, the Analog Agentic Workbench: an AI-assisted environment being developed to connect engineering requirements, specialized agents, circuit simulations, and design decisions in one traceable workflow.
Building the future of analog EDA•Early-stage prototype development
WORKFLOW CANVAS
INPUTRequirementsSpecs & constraints
AI AGENTDesign assistantTopology exploration
EDA TOOLSimulationRun & measure
HUMAN GATEEngineer reviewDecision & sign-off
Conceptual workflowDAG • traceable execution
FIG / 01
01 / The challenge
Engineering complexity shouldn't mean workflow chaos.
Analog circuit development spans specifications, design choices, tool runs, and verification cycles. Information and decisions often live across disconnected systems.
Costly iteration
A small change may trigger manual checks and repeated runs without clear visibility into what actually changed.
Missing context
AI suggestions need provenance, tool results, and human engineering judgment to become useful in real workflows.
02 / The approach
One connected journey from spec to evidence.
ANVIL's proposed workbench brings specialist reasoning and deterministic engineering tools into a human-supervised orchestration layer.
Example analog design flowILLUSTRATIVE • NOT A LIVE PRODUCT DEMO
STEP 01 — STRUCTURED INPUT
Turn engineering requirements into reusable context.
Capture measurable design goals, operating conditions, and constraints in a format future tools and agents can reference.
analog_spec.yaml ● DRAFT
design: op_amp
targets:
gain_db: ">= 60"
bandwidth: "project-defined"
status: review_requiredSTEP 02 — SPECIALIST AGENTS
Coordinate the right expertise for each decision.
Route work between requirements, topology, sizing, and verification roles. The engineer retains control over proposed next steps.
STEP 03 — TOOL EXECUTION
Ground suggestions in simulation evidence.
Connect authorized tools and simulators through auditable adapters; collect outputs, measurements, and execution metadata.
STEP 04 — HUMAN OVERSIGHT
Put the designer at the approval boundary.
Proposed human gates allow design owners to inspect trade-offs before committing changes or advancing to later phases.
STEP 05 — CHANGE-AWARE ITERATION
Keep the why alongside the what.
Track artifact dependencies and design history so changes can guide targeted re-evaluation instead of blind repetition.
03 / Design principles
Built around the engineer, not instead of one.
These principles guide the planned architecture and experimental roadmap; they are not a claim that every capability is implemented today.
Specialized AI agents
Focused roles for parsing specifications, considering topology options, interpreting results, and assembling review context.
Tool-first execution
Explicit adapters between reasoning and simulator/EDA tools, with deterministic outputs where possible.
Visual orchestration
Composable workflows with agent, tool, and approval nodes, designed for parallelism and dependency awareness.
Traceability by default
Attach rationale, inputs, artifacts, and execution evidence to engineering decisions and iterations.
PLANNED INTEGRATION MODEL
Modular • human-supervisedWorkspace & orchestration → agent/tool adapters → authorized simulators → evidence & review
04 / Current status
Research-led. Prototype-focused.
ANVIL is an early-stage, five-person engineering initiative developing an AI-powered platform for analog IC design and verification.
We are currently building and validating our initial prototype, focusing on intelligent agent orchestration, EDA tool integration, and reproducible engineering workflows.
Our long-term goal is to bring ANVIL to market as a platform that helps engineering teams accelerate and streamline analog design workflows.
Foundations & integrations
Specify the workflow schema, tool adapters, and a reproducible example design flow.
Orchestration & agents
Add role-based assistance, execution control, and persistent artifact tracking.
Engineering validation
Measure correctness, traceability, runtime, and iterative productivity on suitable benchmarks.
Tool integrations, including commercial EDA platforms, depend on authorized access, licensing, and compatibility testing. ANVIL does not claim affiliation with or endorsement by EDA vendors.
LET'S CONNECT
Email ANVIL Better engineering workflows start with better conversations.
We welcome discussions with researchers, analog designers, mentors, and potential collaborators interested in AI-assisted engineering automation.